| # Example GitHub Actions job: block a PR when the candidate RAG config is worse than the baseline | |
| # by more than the margin, using the paired non-inferiority gate (G3) from this project. | |
| # Assumes your eval step writes per-query scores to CSV files with columns: qid,recall@5 | |
| name: rag-gate | |
| on: [pull_request] | |
| jobs: | |
| gate: | |
| runs-on: ubuntu-latest | |
| steps: | |
| - uses: actions/checkout@v4 | |
| - uses: astral-sh/setup-uv@v6 | |
| - run: uv sync | |
| # Replace with your own evaluation; both files must cover the same golden-set queries. | |
| - run: uv run python your_eval.py --config main.yaml --out baseline.csv | |
| - run: uv run python your_eval.py --config candidate.yaml --out candidate.csv | |
| - run: uv run python -m gatestats.gates baseline.csv candidate.csv --metric recall@5 --margin 0.02 | |